Anda di halaman 1dari 31

The manufacture of synthetic fibers represents a huge industry, both in the United

States and worldwide. In 1990, the dollar value of synthetic fibers was roughly $17
billion in the U.S. and $70 billion worldwide. Natural fibers, consisting mainly of
cotton and wool, but also including silk, yielded total dollar values only slightly
higher.
We sometimes forget how much of these fibers we consume, especially since the
appeal of "natural" fibers of cotton and wool has grown in recent years. Despite
questions of aesthetics and taste, there should continue to be a significant demand for
synthetic fibers, in large part because these fibers can be tailor-made to provide
specific properties that natural fibers cannot provide.
As chemical engineers, we are interested in the manufacture of synthetic fibers
because we have been the principal developers of the processes used to produce the
fibers and because we are usually the ones charged with overseeing and improving the
manufacturing operations.
As students of chemical engineering, we are interested in the manufacture of synthetic
fibers because the process involves several fundamental aspects of chemical
engineering. Once we understand this process thoroughly, we will be in a position to
understand other chemical processes quite easily. The goal here is to provide us
students with the motivation and curiosity to learn these fundamental concepts; such
motivation is supposed to stem from exposure to practical and somewhat familiar
operations where the value of learning the fundamentals will become self-evident.
Figure 1.1 depicts the four main areas of fiber manufacture: pumping, filtration, fiber
forming, and fiber treatment. We have purposely excluded the production of the
polymer melt or solution, so that we could focus on those fundamentals associated
with the transport phenomena, rather than the chemistry, of fiber production. We also
point out that synthetic fibers include not only the familiar examples of nylon or
polyester, but also could extend to fiber optic cables and wire, etc.



Table 1.1 depicts the principal elements of transport phenomena fundamentals which
are either described or might be included in the detailed study of the process of
synthetic fiber manufacturing. Illustrated in Table 1.1 are the principal elements
covered in the module which also might be seen in the table of contents of a textbook
on transport phenomena. Particularly noteworthy is the breadth of topics included in
this one module, ranging from pumps and filtration to transport analogies for the three
modes. The student may be less interested in this particular table than the teacher, who
is often concerned with extent of coverage. Clearly, the main goal at the present is to
preserve the student's interest and curiosity by emphasizing the relevance of the
topics.





Table 1.1 Transport Phenomena and Mathematics Fundamentals


In Figure 1.2, we attempt to provide the student with some orientation and
introduction to the glossary of fiber production.


Fabric This includes a description of the difference between non-woven and woven
fabrics. Whereas woven fabrics have an orientation associated with them, in which the
fibers are normally aligned either parallel or perpendicular to each other, non-woven
fabrics do not possess any preferred fiber orientation. With synthetic fibers produced
from molten polymer, non-woven fabrics can be bonded together by having the
filaments laid down over one another while they are still molten. As they cool and
solidify, they are "glued" together. The fibers can be laid down in much the same way
that paint appears to be laid down in a Jackson Pollock painting!

Figure 1.2 Synthetic Fibers - A Little Taxonomy
Fibers Fibers can be either natural or synthetic and either continuous or staple. Natural
fibers, of course, can come from either animals or plants and probably the most well-
known example of each are wool and cotton, respectively. Their chemical structure is
polymer-based, in that a regular, repeat structure can be found in natural fibers.
Synthetic fibers, too, are based on a regular polymeric structure. However, synthetic
fibers are manufactured, or "synthesized," usually from oil, but sometimes from coal
or natural gas. Most of the synthetic fibers are from polymers produced by step
polymerization (usually, condensation polymers) but many are made be from
polymers produced by chain polymerization (addition polymers).
All natural fibers have a finite length associated with them, ranging from about 5 to 20
cm. This could be the length of the hair on the sheep (wool) or the length of a cotton
filament in a cotton plant. In order to be woven into a fabric, these filaments must first
be aligned together into a continuous strand, called yarn or thread. Spinning machines
are used to accomplish this. The filaments are held together by van der Waals' type
forces. The thread and yarn produced in this way contain occasional filaments which
stick out away from the continuous strand; this is somewhat like branch groups on a
polymer, albeit at a much larger scale. In any case, these filament branches help to
provide the woven fabric with greater bulk and porosity and are associated with other
positive aesthetic features of the fabric. Synthetic fibers, however, can be made into
continuous filaments which are practically infinite in length. Even though the
individual polymer molecules in the filament are 1000 x longer than they are wide,
they are still usually tiny fractions of millimeters in length. A "yarn" can be made of
these filaments simply by bringing the filaments together continuously as the
filaments are produced. This will not produce a yarn with filaments which
occasionally stick out from the strand, however. In order to produce such a result, in
some processes the continuous strands are chopped into strands of finite length, so
that there is a somewhat closer match between the synthetic and the natural fibers.
These chopped-up sections are called "staple," and the staple is brought together again
in spinning machines which operate just like those used with natural fiber. Another
difference between natural and synthetic fibers is that the natural ones are usually
more curly. Two techniques have been used to make the synthetic fibers and yarn
more curly: one is to pass the yarn between two heated gears which can impart a
permanent crimp to the yarn and the second is to produce the synthetic filaments from
two different polymers passing out of a common hole, or die. The resulting filament
will exhibit varying curliness, based on humidity conditions if the two polymers
absorb water to differing degrees and if the absorption changes the shape of the
filaments accordingly.
Synthetic Fibers This list brings together most of the fibers that the with which the
students are already familiar. Most are made from condensation polymers, although
acrylic is not, and carbon fibers are made from addition polymers subjected to a
pyrolysis step to kick off the hydrogen atoms.
Fiber Forming This section includes a brief forecast of what will be covered
subsequently. In particular the three principal fiber-forming processes of melt, dry,
and wet spinning are described briefly. All three steps involve the formation of
continuous filament strands by forcing the material through circular dies, but melt
spinning involves cooling of the subsequent strand to form the solid filament, whereas
dry and wet-spinning involves removal of a solvent to form the solid filament. In dry
spinning the solvent evaporates into a gas and in wet spinning the solvent is leached
into a liquid bath.

Figure 1.3 shows a schematic of the melt spinning process, and thus illustrates the key
elements listed in Figure 1.1. The molten polymer (in the case of dry or wet spinning,
the spin dope) is first pumped through a filter which removes any tiny particles that
can be trapped in the tiny spinneret holes. The polymer is then forced through these
tiny holes to form continuous strands of polymer filaments, or synthetic fiber. Cooling
gases reduce the temperature of the filaments so that they solidify and an initial drive
roll controls the initial take-up speed. The fiber may undergo subsequent heating and
stretching to impart additional molecular orientation. Finally, the fiber is taken up
onto bobbins at a constant speed, with a special tension control device to control the
rate of rotation in order to maintain constant yarn speed.


In this section, we introduce two of the main devices used to pump liquids:
centrifugal pumps and gear pumps. The centrifugal pumps are general-purpose
devices used to move low-viscosity liquids around in a process; gear pumps are used
to pump highly-viscous liquids at a controlled flowrate. In order to form synthetic
fibers, a viscous liquid must be pumped through a filter bed and then through tiny
spinneret holes, and both units will result in high pressure drops along the flow
direction of the viscous liquids. Finally, the flowrate must be maintained within close
tolerances in order to produce synthetic fibers of uniform diameter. Gear pumps are
used to provide such a closely regulated flow of the highly viscous liquid; although
gear pumps could also be used to develop the high pressures required to force the
material through the filter and the spinneret, it is more common to use screw
extruders for this purpose.


Figure 2.1. Sketch of centrifugal pump.

Perhaps the best way to understand the way centrifugal pumps work is to recall the
mechanical energy balance equation. (In the present analysis, this will be Bernoulli's
equation, with the additional term of shaft, or pump, work included.)
The mechanical energy balance (MEB) equation is:

(5-1)

Recall that this equation is valid, under steady-state conditions, along a streamline of
the liquid. Since we are dealing with a liquid of relatively low viscosity, we can
ignore the viscous term, lwf, and, with constant density, the integral term simplifies,
so the modified MEB equation (Bernoulli's equation with shaft work included)
becomes:
(5-2)

Referring to Figure 2.1, we see what amounts to a spiral fan, with the inlet at the fan
axis and the outlet in a line tangential to the tip of the rotating fan. The fan, or
impeller, is usually of uniform thickness or width and is sandwiched in a casing
formed by two parallel disks which are sealed at the edges with a short, squat cylinder
with one exit tangential to the cylinder. In operation, the liquid enters along the axis,
into the "eye" of the impeller and both the pressure and velocity are relatively low at
this entrance. As the liquid travels through the pump, the rotating impeller
(previously, spiral fan) imparts shaft work to the liquid as the liquid develops higher,
mainly -direction, velocity. As a matter of fact, within the pump,
v = 0 er + (Vr)(r/R) eq + 0 ez (5-3)

where VR is the -direction speed of the liquid at the tip of the impeller (which should
only be slightly less than the actual tip speed, depending on the design of the pump)
and er, eq, and ez are the orthogonal unit vectors in the r, q, and z directions,
respectively. The reason that v varies linearly with r/R is geometric; think of the
example of a rock attached to a string as you swing the rock around your head-the -
direction velocity of the string at any radial position varies as r/R. This means that the
amount of motor power delivered to the impeller is concentrated towards the outer
parts of the impeller, because the shaft work is proportional to the square of the
velocity, from Equation (5-1). Getting the liquid out to a radial position halfway
towards the exit requires only one-fourth of the power and two-thirds out still less
than half (only 4/9) of the power. At the tip of the rotating impeller, most of the pump
power or shaft work has gone into increasing the velocity of the liquid; the MEB
equation shows that this arises in the kinetic energy term. If the liquid encounters no
resistance downstream of the pump-let's say that it goes directly into a vertical
fountain with no additional change in cross-sectional area of the exit tube-then very
little pressure is "developed" by the pump. Again, pursuing the fountain example, you
will recall calculations in physics about how the height reached by a ball thrown into
the air can be calculated from the conversion of kinetic energy at the bottom to
potential energy at the top, and you realize that the same thing could be done here.
You could, theoretically, calculate the impeller tip speed from the height of the
vertical fountain. Of course, you quickly realize that problems with friction inside the
pump, changing tube sizes, and friction with the air would likely render the
calculation too inaccurate. But you do realize that the same principle can be applied
here. In the typical setting, the fluid does encounter resistance downstream of the
pump. In the extreme, the resistance is infinite (the valve is closed!) and now the
pump is used to develop pressure, instead of velocity. If we now apply the MEB
equation from the impeller entrance to a point just past the rotating impeller, where
the liquid velocity is zero, the change in kinetic energy, KE, is zero, but the change in
potential energy, P/r is equal to the shaft work, Ws (per unit mass of liquid pumped).
Before we move on to some of the design equations with centrifugal pumps, we
should briefly mention some other pumps for low-viscosity liquids:
Straight-through pumps: Axial fan and turbine pumps are typical of these units,
although they are more commonly used with gases.
Piston/cylinder pumps: Multi-stroke units are used as metering pumps, where a
constant flowrate, though pulsating is needed; single-stroke units are typified by
infusion pumps, which are useful in medical applications where medicine has to
titered into a patient over a sufficiently long time period that replenishing the stroke
volume produces no problem.
Air-lift pumps: These produce vertical, two-phase flow and, although they are not
energy efficient, are relatively cheap and can be used with highly corrosive liquids.
Drip coffee makers work on this principle, moving the hot water to the top by
vaporizing a portion of the water and the resulting steam bubbles rise in a tube,
dragging and pushing the hot water upwards


Figure 2.2. Sketch of section of a gear pump.

These pumps are also aptly named, since they consist of two intermeshing gears. One
intriguing part of these pumps is that, for most people, the flow goes in a direction
different from that first thought. Referring to the sketch on the diagram, we see two
rotating gears, set in a Figure-Eight chamber, with the fluid being carried around the
outside periphery of each gear before meeting again to be pumped away. As a matter
of fact, the fluid is carried in discrete packets as it moves towards the exit. Each
packet moves at a speed directly proportional to the rotational speed of each gear, so
the total volumetric flowrate is proportional to the rotational speed of the pump.
Although the gear teeth are shown in the diagram as square-toothed, they can also be
sinusoidal. If the gears are sinusoidal, then the output flowrate from one gear will vary
sinusoidally, from zero to a maximum value. Corresponding output from the other
gear will also be sinusoidal, but the phase difference will be exactly radians. This
means that sum of the two outputs will be a steady, or uniform, flow because the two
flows will exactly balance each other. While this is not exactly like destructive wave
interference learned in physics in the study of light and sound waves, since we have a
uniform maximum output remaining, it does appear similar.
In fiber spinning operations, the gears are normally square-toothed, with a relatively
large number, say 40 or so, teeth per gear. The gears are usually designed to operate
over a relatively narrow range of rotational speeds and many utilize a standard
diameter, so additional capacity, or flowrate, is obtained by choosing gear units of
varying thickness. This would be analogous to centrifugal pumps being designed for
higher flowrates by using impellers of greater thickness. For other applications, such
as pumps for heavy oils and lubricants, one normally finds fewer teeth per gear
(perhaps eight or so) and a sinusoidal shape to the teeth.
With gear pumps, the clearance between gear teeth and barrel of gear housing is
important, because, with an adverse pressure gradient, the liquid can leak back
through the pump through this clearance. This is one of the reasons that gear pumps
used in fiber spinning, which may be exposed to opposing pressures of a 1000 psi, are
square-toothed. The flow between the teeth tips and the barrel is nearly Couette,
which the student will recall as flow between parallel plates. One can calculate the
amount of leakage by taking into account the difference between the drag flow,
produced as one moving parallel produces a linear velocity profile and flow in the
pumping direction, and the adverse pumping flow, which produces a parabolic type of
flow in the opposite direction. This difference can be shown to depend strongly on
this distance between the teeth tips and the barrel.
Starting up a spinning line is a complex operation and one wants to minimize the
frequency. It also is time-consuming to replace spinneret plates and filter beds.
Inclusion of particles, whether they are thermally degraded polymer or small chips of
metal, can lead to imperfections along the filament and potential points where the
filament may break. Worse still, the particle may become attached to the entrance to a
tiny spinneret hole and this could reduce the flow to that hole and thus the diameter of
the resulting filament; this smaller filament will cause the yarn to take up dye
differently and the resulting non uniformity will show up in the final fabric. Another
problem with partial blockage of flow is that the blockage may be nearly complete,
leading to breaking of the filament, which is usually only a very slight problem, since
there are so many filaments, although gradual oozing of the melt through the nearly
blocked hole can be a very major problem. It becomes a major problem because the
polymer, as it oozes out the hole, will gradually travel to neighboring holes and
interfere with the filaments from them as well. The only solution for this problem is to
shut down the line and remove and clean the spinneret plate.
Since there can be up to 1000 holes in a spinneret plate, the large number means that
the filtration process must be completed to very exacting standards. Similarly, because
one wants to operate the line for long periods of time between shutdowns, one wants
to use a filter medium which can trap a large number of particles without causing
excessive pressure drop or blinding of the filter. This brings us to the first part of
filtration, deciding between filter screening and entrapment among small particles.
Before the mid 1960's, draft beer was sold in large containers, called kegs, under
refrigerated conditions. Draft beer, which had not gone through a high temperature
pasteurization step to destroy the yeast and any bacteria, had to be consumed within a
short time of its brewing and had to be kept refrigerated prior to consumption, in order
to avoid problems with further fermentation. Purists liked the draft beer because they
claimed that the pasteurization step, which was necessary for beer sold in small
containers at room temperature, would destroy some of the delicate flavors of the
beer. In the 1960's, the beer manufacturers found that one could substitute an
extremely fine filtration process for the pasteurization step, since all the yeast and
bacteria could be simply filtered from the beer. The relatively small molecular species
producing the delicate flavors could pass through the very tiny filter holes. The
resulting beer could be packaged under sterile conditions and could then be sold in
cans or bottles stored at room temperature. This is an example of very fine screen
filtration, where the requirements for filtration of the yeast and bacteria could be
carefully determined. Of course, the yeast do tend to pile up on the filter screen and
blind it periodically, requiring replacement of the filter screen.



Screen filtration can be accomplished under batch or continuous operation. With batch
filtration, as described above for the beer example, the operation can be conducted
under a) constant pressure, in which the volumetric flowrate decreases with time or b)
constant flowrate, in which the pressure drop increases with time, as the filter
becomes loaded with particles. Commercially, one frequently encounters operations
requiring the filtration of particles which are rigid and incompressible, unlike bacteria.
With these liquid/particle systems, the particles build up on the screen, forming a filter
cake. If the resulting filter cake is indeed incompressible, equations (from Momentum,
Heat, and Mass Transfer, 3rd ed., by C. O. Bennett and J. E. Myers, McGraw-Hill,
New York, (1982), p. 233) for the filtration process are:
(3-1)

where
f
is the filtration time, V
f
is the volume of filtrate, P is the pressure difference
(usually upstream gage pressure for atmospheric filtration) and K
1
and K
2
are
characteristics of the filtration, suspension concentration, and particle size and shape.
K
1
and K
2
would normally be determined by experimentation, although Bennett and
Myers do provide equations to relate the values to such variables as specific cake
resistance, concentration, and viscosity of fluid.
For filtration at a constant volumetric rate q
0
, the equation is:
(3-2)

where K
1
and K
2
are the same as for constant pressure filtration. Thus, one can predict
constant flowrate filtration performance from constant pressure data and vice versa.
All of this is predicated on incompressible filter beds, however, and at least half, if not
more, commercial slurries exhibit compressible bed performance. Although one can
attempt to account for this by a dependence of specific cake resistance on pressure,
the final design normally requires close experimental verification anyway.
With deep-bed filtration, illustrated by Fig. 3-2, the particles to be captured are
entrapped in the interstices among particles held in a deep bed. The sizes of the
interstices are roughly the same order as the size of the particles in the bed, so, in
order to maximize the capacity of the bed, the particles are laid down in such a way
that a gradient of particle sizes exists in the bed. Larger bed particles at the top, or
beginning, of the bed will tend to trap the larger particles and smaller bed particles at
the bottom of the bed will trap the smaller ones. The example on the sketch shows a
gradation in sand particle sizes produced by taking the sand from different locations
along the eastern seaboard. In practice, although sand is frequently used in such deep-
bed filters, it usually is river bed sand and not salt-water sand, because the former
tends to be more free of extraneous materials.



The pressure drop for laminar flow (which will certainly hold for molten polymer
melts and polymer spin dopes) can be calculated from the mechanical energy balance
equation, MEB, and the Kozeny-Carman equation. The MEB (from Bennett & Myers)
is written as:
(3-3)
As mentioned previously, fibers are formed by the extrusion of the polymer melt or
spin dope through tiny holes in a spinneret plate. Such a plate may contain 1,000 holes
or more. Textile fibers are relatively fine, so the diameter of the hole may be only a
few mils, where one mil is 0.001 inches (25.4 µm). The thickness of the
filament is generally not given in linear dimensions, but rather in terms of mass per
length. For some reason the fiber industry has adopted the terms denier and denier per
filament, dpf, to express the filament size. One dpf corresponds to a mass of 1 g in a
length of 9000 meters! If the density of the polymer is 1 g/cm3, this would correspond
to a diameter of 1.2 x 10
-3
cm, or about half a mil. Typically, textile fibers are in the
range of 3 to 15 dpf. Recall that one g is roughly 1/30 of an ounce.
In melt spinning, the filaments are normally drawn down, or stretched, just
downstream of the spinneret holes. The stretch is of the order of 2 to 3x, so the
spinneret hole may be 50 to 75% larger than the filament diameter when it is first
cooled. Additional post-formation stretching may also be used, however, so that the
final filament diameter may be one-half or less than the diameter of the spinneret hole.
The spinneret hole is usually only slightly longer than its width, in part to minimize
pressure drop at the plate. But the plate still has to be strong enough to withstand the
upstream pressure. For this reason, the melt passes through a conical section before
reaching the final spinneret hole, so that the plate can be relative thick (see Fig. 4.1).
Pressure drop through this converging section is very difficult to calculate for these
polymeric materials, because the extensional flow rheology is usually not well
characterized. One can readily visualize the alignment of the polymer molecules in the
converging section, where the polymer undergoes a severe stretching step. Ignoring
this pressure loss, let us focus on the spinneret hole itself.
An important question here is whether one can use the Hagen-Poiseuille equation to
compute the pressure drop in such a short tube. To help answer this question, we first
compute the entrance length, which is approximately equal to the axial distance
downstream from a tube entrance at which the momentum boundary layers merge at
the center axis, where a fully parabolic profile is established. This distance, from Bird,
Stewart, and Lightfoot, Transport Phenomena, p. 47, is
Le/D = 0.035 Re (4-1)
where Le is the entrance length, D is the tube diameter and Re is the Reynolds
number.





For a representative calculation, we consider nylon melt, with a viscosity of 200
Poises, being spun from a hole 10 mils in diameter at a final spinning speed of 2,000
ypm and a stretch of 5x between the spinneret and the final take-up. This means that
the bulk velocity, ub, in the spinneret hole is 400 ypm. The Reynolds number, for a
specific gravity of about one, is:

Re = ub D/n = 0.077 (4-2)

so the entrance length calculated from Eqn. 4.1 is less than 3 thousandths of the
diameter of the hole. Therefore, we can safely use the Hagen-Poiseuille equation to
calculate the pressure drop. The equation is:
(4-3)

For the nylon example we just explored, the pressure drop is predicted, for a length of
3.0 mils, to be 2200 psi. This pressure drop might be a bit excessive in practice, but
the method of calculation remains illustrative. One part of the calculation which was
not taken into account is the power-law behavior of most polymer melts and polymer
solutions. Such behavior usually is revealed by a shear-thinning response, in which
the apparent viscosity decreases as shear rates increase. This would lead to
significantly lower pressure drops for the spinneret plate.
As the polymer exits the spinneret hole, it tends to swell and this swell is especially
noticeable at low filament tensions. Apparently, the polymer molecules must coil
under the shearing action within the hole and, as it exits, the polymer molecules are
free to uncoil, as seen by an expansion of the polymer stream jetting out of the hole.
This phenomenon is referred to as "die swell," and can even amount to a doubling or
more of stream diameter. Newtonian fluids can also be shown to exhibit a swelling at
the exits of tubes, even at very low Re; the predicted extent of swell is about 14% for
Newtonian fluids. Since, in fiber spinning, the filaments are under tension, the extent
of die swell is considerably reduced. Furthermore, the extent of swell appears to have
no influence of final filament properties. In order for the filaments to undergo
stretching, some power must go into the stretching motion immediately downstream
of the spinneret plate, but the amount of this power is negligible, as described in the
next section.
In the spinning of molten polymers, such as nylon, polyester, and polypropylene, melt
spinning begins with a cooling of the molten filament after it leaves the spinneret. At
the same time, the filament is pulled downwards towards the take-up section and this
resulting tension in the molten filament provides a stretching action in the molten
filament itself. In most melt spinning operations the degree of stretch is of the order of
3x, which means that the velocity of the initially cooled, or solid, fiber is about three
times the average velocity of the melt coming out of the spinneret. For some
filaments, this initial stretch is very important in helping to establish properties in the
polymer which depend on whether one deals with the properties in the fiber axis
direction or in the fiber radius direction. This directional dependence of properties is
called anisotropy and the usual example is that of a slab of wood, in which strength
and fracture properties along the grain are quite different from properties across the
grain. (With many fibers, however, these properties are controlled downstream, where
the fibers are reheated, stretched further, and cooled again.)
In any case, the polymer melt, once it comes out of the spinneret hole, starts to cool
down and also starts to stretch out. Because the "apparent viscosity" of the melt
increases rapidly as the melt cools, most of the stretching takes place in a region
relatively close to the spinneret hole whereas "most" of the cooling takes place well
away from this hole. But these terms and descriptions are not exact and are not easily
quantified. The real advantage in using these descriptions is that it permits us to make
a simplifying assumption as we analyze the melt spinning process. The assumption is:
We can separate the stretching and cooling operations into two separate distinct
regions, with the first occurring relative close to the spinneret (say, within 10% of the
distance to the first take-up, or speed-control roll), and the second over the remaining
distance to the first take-up roll. If need be, we could return later to this assumption to
determine its degree of accuracy, but let us accept it for the moment.
The stretching region, within which the relatively long polymer molecules become
aligned along the filament axis, might be characterized by very complex rheology.
Within the field of polymer processing, rheology deals with the relationship between
stress and the history of strain; for Newtonian fluids, you can recall that the fluid
stress is proportional to the instantaneous rate of strain (the shear rate). We are not
really too concerned with the polymer melt rheology here, however, since it will not
likely be important in determining the power required for the first drive or take-up
roll. Frictional and interfacial stresses are likely to be far more important. Therefore,
in terms of design considerations, we can probably ignore that part of the melt-
spinning process in which the initial, post-spinneret stretching of the polymer melt
occurs and focus instead on the cooling step of the melt spinning operation.



Perhaps the most important design consideration in the melt spinning process is the
cooling of the filaments, Fig. 4-2. In order to simplify our analysis, we restrict our
focus to the point where the filaments have reached a uniform diameter (recall that we
previously asserted that this is a relatively short distance) and that they are at some
initial temperature which will be somewhat cooler (by about 20 C°) than the melt
temperature at the spinneret exit. At this point, the temperature within the filament
will depend on radial position, with the maximum occurring at the center, on the
filament axis. We shall invoke an approximation of a flat temperature profile, in
which the temperature does not vary with r at this intial position, in order to utilize
existing mathematical solutions. An important part of learning engineering is to learn
how to take "appropriate shortcuts" which save time with little sacrifice in accuracy.
This is one example. The melt spinning process is steady: viewing the spinning
threadlines at a fixed position (the socalled Eulerian perspective) shows that nothing
appears to change with time. If one situates oneself on the moving threadline
(figuratively, of course) there does certainly appear to be a time dependence to the
temperature of the filament. This viewpoint of moving with the material is called the
Lagrangian perspective. Whereas the Eulerian perspective requires you to measure the
threadline temperature as a function of r, radial position within the filament, and z,
axial position along the filament, in order to follow the cooling, the Lagrangian
perspective allows you to follow the cooling as a function of r and t, where t is time.
Zero time should be some convenient reference-here it would correspond to locating
yourself on the filament at the end of the stretching region and at the "beginning" of
the cooling region. This is equivalent to the cooling of an infinite rod which is fixed in
space. The governing differential equation, which can be derived easily using shell
balance techniques is:
(4-4)

where is time and is the thermal diffusivity of the polymer. The student will readily
recognize this as a partial differential equation, since the temperature T depends on
both r and . In order to solve the equation quantitatively, one must specify initial and
boundary conditions. The boundary is naturally R the outside radius of the filament.
So the initial condition ( = 0) is simply:
T (r, ) = T
o
for r< R (4-5)

where T
o
is a constant. We need two boundary conditions, corresponding to r = 0 and
r = R. At r = 0,
T remains finite; although some prefer to say:
(4-6)
based on symmetry arguments. At r = R, the heat arriving at the surface by conduction
from within must match the heat leaving by convection:
(4-7)
at r = R and all > 0. k is the thermal conductivity of the filament (we shall assume
that this conductivity does not change as the polymer solidifies. h is the heat transfer
coefficient governing the heat transfer from the surface to the surrounding air. h can
be estimated from various correlations if information about the velocity and direction
of the cooling air is given. An example of such a correlation for heat transfer from a
cylinder in crossflow is given by Churchill and Bernstein (J. Heat Transfer, 99, 300
(1977)):
(4-8)

where NuD is the Nusselt number, hD/k, (k is the thermal conductivity of the fluid in
crossflow and D is the cylinder diameter) and ReD is the Reynolds number based on
the fluid in crossflow. Pr is the Prandtl number, v/ ,and is also based on the crossflow
fluid. The student will recall that an important advantage of presenting correlations in
terms of dimensionless variables like Nu (dimensionless heat transfer coefficient) and
Re (inertial stresses divided by viscous stresses) is that the resulting expression is
often simpler, revealing more clearly the relationships among such variables.
We shall also assume that the heat of fusion is negligible, primarily because we want
to simplify the calculation. This assumption, especially for crystalline polymers, could
be very poor, however, and could lead an underestimate of the cooling time by a
factor of two.
As engineers, or even as normal, sane people, we would not want to solve the
differential equation for every single set of geometries, thermal diffusivities and initial
and boundary conditions. We can avoid this needless energy expenditure if we
express the differential equation in dimensionless form:
(4-9)
where:
(4-10)

Y is called the unaccomplished temperature change, since it starts at unity at time zero
and declines from there. n is the normalized radial position, and X is dimensionless
time, sometimes called the Fourier number. One final dimensionless group, m, (also
equal to 1/Bi) expresses the relative resistance outside the filament to that within the
filament:
(4-11)

Finally, the initial and boundary conditions become:
Y = 1 at X = 0 and 0 < n < 1; (4-12)

at n = 0 and X > 0; (4-13)

at n = 1 and X > 0; (4-14)

The solution, Y(n, X) is then valid for any case of unsteady-state heat conduction
within a cylindrical geometry with a uniform initial temperature and convective heat
transfer from the surface to a surrounding fluid at a uniform temperature To . The
solution is shown in graphical form on the slide and is available in almost all transport
textbooks. The resulting charts are known variously as "Gurney-Lurie Charts" or
"Heissler Charts," depending on which reference or form of charts you use. An
analytical solution for a slightly less general case is given below. Note that, by use of
dimensionless variables, we have successfully created a result which is applicable to a
broad range of geometries and material properties. For the special case in which heat
transfer resistance from the surface of the fluid to the surrounding fluid is negligible,
one can set h, the heat transfer coefficient, to infinty (this, of course, is equivalent to
setting the temperature of that surface to that of the surrounding fluid for all > 0) and
the analytical solution is:
(4-15)

where J0 is the zero order Bessel function of the first kind and ai is the ith root of
J0(ai) = 0. This solution is presented in the transport text Momentum, Heat, and Mass
Transfer, by Bennett and Myers, 3rd ed., p. 286.
A solution for nonzero m, or finite heat transfer resistance to the crossflow fluid, is
(4-16)
where J0 and J1 are Bessel functions of the first kind (zero and first order,
respectively), and ai is the ith root of aiJ1(ai)/J0(ai) = Bi.
Key elements that you have learned in this section include:
Difference and relationship between Lagrangian and Eulerian perspective,
Existence of unsteady-state heat conduction charts,
Heat transfer resistance within objects relative to resistance outside these objects,
Value of dedimensionalization as a means to obtain more general solutions to
complex equations,
A typical heat transfer correlation for heat transfer coefficient, h
Spinning of synthetic fibers with a melt spinning process is conceptually simple,
involving little more that the extrusion of molten polymer through fine holes and
solidification of the resulting filaments by cooling

Unlike melt spinning, both dry and wet spinning use solvents in which the polymer
dissolves, see Fig. 4-3. The resulting solution or suspension is a viscous "spin dope."
This process necessarily introduces another species, which is subsequently removed,
and therefore is more expensive than conventional melt spinning processes. It is used
in cases where the polymer may degrade thermally if attempts to melt it are used or in
cases where certain surface characteristics of the filaments are desired-melt spinning
produces filaments with smooth surfaces and dry spinning produces filaments with
rough surfaces. The rougher surface may be desirable for improved dyeing steps or for
special yarn characteristics.



Fibers produced by wet spinning include rayon and Kevlar. Rayon was originally
developed as a synthetic substitute for silk and Kevlar was produced as a high-
strength fiber for use in various aerospace and specialty-use applications.
Furthermore, many comercial acrylic fibers are also produced by wet spinning.



As with dry spinning, the polymer is dissolved or suspended in a solvent, to form a
viscous "spin dope" and filaments are formed by extrusion through tiny holes in a
spinneret plate. Kevlar, for example, will degrade thermally if attempts are made to
melt it, and thus a solvent must be used. The term wet spinning more accurately
depicts the process than does dry spinning, because the solvent is extracted or,
perhaps more appropriately, leached, from the filaments by another liquid. In most
cases, the second liquid is aqueous.
A major difference between wet spinning and either melt or dry spinning is that one is
spinning into a fluid (liquid) with a much higher viscosity. Because this higher
viscosity can translate into high shearing stresses on the surfaces of the filaments, the
tension in the filaments can become quite high. For example, towing a buoy by a long
line behind a boat can produce very high tensions in the line when compared with
towing the same buoy by a short line. For long baths, the tension can become
sufficiently high that the filaments might break, as their tensile strength is exceeded.
To avoid this danger, much lower spinning speeds must be used. Whereas melt
spinning may utilize spinning speeds of 2,000 yards per minute (80 mph), spinning
speeds in wet spinning are usually less than 300 ypm.
Another difference with dry spinning is the capability of using many more spinneret
holes in the case of wet spinning. The total number can approach 60,000 in a single
spinneret plate, if the spinning is done directly into a coagulating or extracting liquid.
Because the liquid is present, the filament forms a type of skin almost immediately
and the potential for the filaments to touch and fuse is practically eliminated,
compared with dry or melt spinning.
In the case of Kevlar the spin dope is relatively warm, about 100&deg;C, and forms a
viscous, liquid crystal. The solvent is sulfuric acid, at a concentration of about 80 wt%
(20 wt% polymer). These liquid crystals are easily oriented by a stretching motion,
but they can lose their orientation, presumably by Brownian motion, once the
stretching is stopped. Therefore, during the spinning process, the filaments are first
extruded through an air gap, where the filaments undergo strains of 2 to 3x, which
produces a high degree of molecular orientation in the filaments, and then they are
suddenly "quenched." This air gap is of the order of one inch. It also allows the
spinneret plate to be warm (100&deg;C) while the extraction bath can be cool (ca
15&deg;C). The hot filaments then strike the cooling bath where the filaments are
quenched and much of the orientation is locked in by the rapid cooling action.
Subsequent to the quench step, the solvent is extracted, which requires a relatively
long bath contact time. But the initial quenching step is crucial, since it allows for the
oriented molecules to be "frozen" into position. This orientation is particularly
important to the high-strength properties of Kevlar-the filaments, on a weight basis,
are significantly stronger that steel. If one attempts to use the same process to produce
Kevlar filaments of large diameter, the core of the filaments can lose its orientation,
because the quench time to reach the core will increase with the square of the filament
radius. The filament skin, or the outer part of the filaments, however, will have the
orientation locked in and a high degree of orientation will exist there. This produces a
so-called "skin-core" effect, in which the average properties of the filaments,
expressed as tensile strength per unit cross-sectional area, will decline on account of a
decreased average orientation.
Kevlar, with its focus on strength development via "air-gap" wet spinning, is
somewhat unique within the process of wet spinning. As with melt- and dry- spinning,
the controlling part of the process is associated with development of the filament
structure, either by cooling of the filament or by removal of the solvent. The equations
for diffusion in wet spinning are identical to those for dry spinning, with the exception
that the fluid passing outside the filaments is a liquid and not a gas. Also, the flow
may not be across the filaments, but even, partially, along the filaments. Therefore,
the correlations and nature of the flow surrounding the filaments will result in
different values for the surface mass transfer coefficient. Whether this will change the
relative resistance dramatically will depend on the particular fiber to be produced and
its dimensions and properties. Note that the same graphical solutions described earlier
can be used.
To design a wet-spinning process, it may be necessary to predict the transport of
momentum, heat, and mass in the region adjacent to the filament just downstream of
the spinneret. One can use a so-called "boundary-layer" analysis to do this. Treatment
of such an analysis is beyond the scope of the present discussion of fiber spinning, but
a brief description of the analysis is appropriate. One form of boundary-layer analysis
involves von Karman integral boundary-layer techniques. The boundary layer starts at
zero thickness at the first point where the fiber contacts the extracting liquid, and
grows gradually radially outwards from each filament as one proceeds downstream.
The velocity profile inside the boundary layer is assumed and all of the velocity
change between the filament and the surrounding fluid is contained within this
"momentum" boundary layer. Similarly, thermal and diffusional boundary layers
contain all the changes in temperature and concentration, respectively. Based upon
approximations of these velocity profiles, frequently assumed to be turbulent, the
variation in filament drag with position can be predicted, along with local heat and
mass transfer coefficients. The student is referred to the text Transport Phenomena, by
Bird, Stewart, and Lightfoot, Wiley, New York (1960) for additional details of such
integral boundary-layer techniques.
Key elements of the wet spinning process include:
Similarity of governing equations to melt and dry spinning,
Lower spinning speeds in wet spinning,
Structure development in air-gap spinning of Kevlar and skin-core effects,
Qualitative introduction to boundary-layer concepts.
In order to start up a fiber spinning line, the filament bundle must be "strung up"
along the process path. The first stage of the spinning process is vertically
downwards, so gravity will tend to pull the line downwards. The line is caught with a
large vacuum nozzle and, almost simultaneously, the line beneath the nozzle is cut
with scissors. The line now is sucked into the nozzle, which functions exactly like a
home cannister style vacuum cleaners, with the fiber piling up in a large drum. The
suction at the nozzle keeps the spinning line under tension. The line is then passed
over pairs of drive rolls, described below, and then over certain stretching/orientation
steps, if needed. Finally, the line is taken to a take-up device which winds the fiber
bundle onto bobbins. In order to maintain a uniform takeup speed, the rotational speed
of the bobbins must decrease as additional fiber is laid on it and the average diameter
increases. This decrease in rotational speed is controlled by keeping the tension
constant in the takeup line. The tension is sensed by deflection of an idler wheel
located between the last drive roll and the take-up machine. The idler wheel is located
on a type of cantilever beam, so that the deflection is a measure of the tension in the
line. The position of the wheel, and thus the line tension, is maintained constant by
adjustment of the take-up speed.
The drive rolls are always characterized by having non-parallel axes. This permits the
threadline to advance in the axial direction on the rolls as the line is strung multiple
times around the rolls. The rolls, which are usually highly polished, permit no slip on
the surface on account of the multiple wraps on each pair of rolls and the tension in
the line. This is somewhat like the calculation of frictional force for a rope wound
around a post, a calculation the student probably performed in a statics course.
Because the drive rolls permit good speed control for each pair of rolls, it is possible,
by having the second pair of drive rolls moving faster than the first, to control the
degree of stretch the fiber undergoes between the two pairs of rolls. This stretch,
which may range from a few percent to 20 or 30%, helps to impart desired molecular
orientation, and thus desired mechanical properties, to the fiber bundle. Sometimes it
is necessary to heat the bundle to a temperature above Tg, the glass transition
temperature, but below the melt temperature, in order to facilitate significant stretches
without imparting excessive line tension. Usually, the strain gives the fiber a higher
modulus (stress over strain), or stiffness, in the axial direction and higher tenacity, or
strength to break per unit cross sectional area.



Table 5.1 summarizes several areas which would be convenient for post-presentation
elaboration, or, perhaps, illustrative asides. For example:
Analytical solution of Partial Differential Equations The diffusion equation is seen
frequently in the physical sciences, and presents a nice starting point for mathematical
topics, including techniques of separation of variables, inhomogeneous boundary
conditions, etc.



Gurney-Lurie & Heissler Charts These charts depict very nicely several physical
results of unsteady-state conduction and diffusion. These include the "nearly
exponential" decline of unaccomplished change with time and, similarly, the
dependence on the square of characteristic dimension, via the Fourier number.
Similarly, the relative resistance effect can be seen clearly, along with quick
calculations to indicate quantitatively, when a lumped-parameter model, where, for
example, internal resistance could be neglected, would be appropriate.
Application to Other Cylindrical Geometries Direct applications could include
glass fiber-optic cable and extrusion of rod-shaped polymer material, although one
could extend the calculations to food processing, such as sterilization times for cans of
food placed in a steam oven, or cooling of cans of soda, etc.
Complication of Accounting for Declining Radius or Heat of Fusion The student
can recall that these complications were removed during our treatment so that we
could proceed with the analysis. One could account for the declining radius by
patching together successive solutions of uniform diameter and uniform temperature
to approximate the smoothly varying geometry with one varying in steps. This would
require knowledge of the extensional rheology as a function of temperature, however.
Similarly, one would have to know the heat of fusion in order to account for its effect.
If the total amount of heat to be removed by fusion is of the same order as that of the
sensible heat, then the Fourier time calculation is likely to be too short by a factor of
two or so.
Extension to other Geometries These would include infinite slabs, for example in
the manufacture of float glass or cooling of metal slabs or ribbons of steel. Extension
of the infinite slabs to Newman's rule so that three-dimensional bricks can be treated
by using three separate one-dimensional solutions. Spheres are also very important,
for example, in the removal of unreacted monomer, VCM (vinyl chloride monomer),
from PVC (poly-vinylchloride) spheres made by suspension polymerization or drying
of porous spheres, and cooling of molten metal pellets, etc.
Convection Currents Inside an Object The time to cool a soda can will be less than
that calculated from the charts because convection currents will become established
with the can that will hasten heat transfer from the can. This is somewhat like the
familiar example from heat transfer courses which states that ice cubes can be made
more quickly from tepid water than from cold water on account of the establishment
of such convection currents, which will more than compensate for the greater amount
of heat to be transferred. The author does not recall seeing any confirming
experimental results, however.